Existential Risk: Assessing Long-Term Humanity Futures Through Convergence

7 gennaio 2026

BY NICOLE LAU

Existential risks—threats that could end human civilization or cause human extinction—are the ultimate high-stakes predictions. Climate change, advanced AI, engineered pandemics, nuclear war. How do we assess risks that have never happened but could be catastrophic?

What if we could assess existential risks using convergence—integrating scientific consensus, technological trajectories, historical precedents, systems modeling, institutional preparedness, public awareness, mitigation efforts, and philosophical frameworks to evaluate which threats are most severe and which interventions are most urgent?

This is where convergence-based existential risk assessment comes in—applying the Predictive Convergence framework to humanity's long-term future, helping researchers, policymakers, and philanthropists prioritize efforts to reduce catastrophic risks.

We'll explore:

Multi-system risk assessment (integrating diverse threat evaluation approaches)
Risk prioritization (using convergence to identify most severe threats)
Mitigation framework (which interventions are most effective)
Case studies (climate change, AI risk, nuclear war, pandemics)

By the end, you'll understand how to apply convergence thinking to existential risk—making better decisions about humanity's long-term survival through multi-system validation.

The Existential Risk Challenge

Why Existential Risk Assessment Is Hard

Problem 1: No historical data

By definition, existential risks haven't happened (we're still here)
Can't use past frequency to predict future probability
Example: Asteroid impact—happened to dinosaurs, but no human experience

Problem 2: Long timescales

Risks may unfold over decades or centuries
Hard to maintain urgency for distant threats
Example: Climate change—slow-moving but potentially catastrophic

Problem 3: Uncertainty and disagreement

Experts disagree on probabilities (AI risk: 5% vs 50%?)
Model uncertainty (climate sensitivity: 2°C vs 5°C?)
Unknown unknowns (risks we haven't identified)

The convergence solution: When multiple independent risk assessment systems converge on high threat, prioritize mitigation; when they diverge, acknowledge uncertainty but don't ignore

Multi-System Existential Risk Assessment Framework

System 1: Scientific Consensus

Expert surveys:

AI safety researchers: Median 5-10% probability of existential catastrophe from AI (varies widely)
Climate scientists: IPCC consensus—warming >3°C would be catastrophic (not extinction, but civilization threat)
Biosecurity experts: Engineered pandemics emerging threat (probability increasing)

Probability estimates:

Toby Ord ("The Precipice"): Total existential risk this century ~1 in 6 (16%)
Breakdown: AI (10%), Engineered pandemics (3%), Nuclear war (1%), Climate (0.1%), Asteroids (0.001%)

Consensus strength:

Climate change: Strong consensus (97% of scientists agree humans causing warming)
AI risk: Weak consensus (wide range of estimates, high disagreement)

Signal: Scientific consensus shows HIGH RISK (strong agreement, high probability) or UNCERTAIN (wide disagreement, low confidence)

System 2: Technological Trajectories

AI capability curves:

Exponential progress (GPT-2 → GPT-3 → GPT-4 → ?)
AGI (Artificial General Intelligence) timeline: Median expert estimate 2040-2060
Risk: Misaligned superintelligent AI could be existential threat

Biotech dual-use risks:

CRISPR, synthetic biology enable creation of novel pathogens
Democratization of biotech (garage labs) increases risk
Example: 1918 flu reconstructed in lab (2005)—proof of concept

Nanotechnology:

"Gray goo" scenario (self-replicating nanobots)—low probability but high impact

Cyber vulnerabilities:

Critical infrastructure (power grids, financial systems) vulnerable to cyber attacks

Signal: Technology trajectories show ACCELERATING RISK (capabilities advancing rapidly) or STABLE (slow progress, manageable)

System 3: Historical Precedents

Near-miss events:

Cuban Missile Crisis (1962)—came close to nuclear war
1983 Soviet false alarm (Stanislav Petrov prevented nuclear launch)
Lesson: We've been lucky, but luck runs out

Past extinctions:

Dinosaurs (asteroid, 66M years ago)—proof that extinction events happen
Megafauna extinctions (humans caused, 10K years ago)—humans can cause extinctions

Civilization collapses:

Rome, Maya, Easter Island—civilizations can collapse
Not extinction, but shows fragility

Pandemic patterns:

Black Death (1347-1353)—killed 30-60% of Europe
Spanish Flu (1918)—killed 50M globally
COVID-19 (2020)—killed 7M+, showed pandemic vulnerability

Signal: Historical precedents show RISK IS REAL (near-misses, past catastrophes) or OVERSTATED (rare events, unlikely to recur)

System 4: Systems Modeling

Climate tipping points:

IPCC models: >2°C warming risks tipping points (ice sheet collapse, Amazon dieback, AMOC shutdown)
Runaway warming scenarios (worst case: 4-5°C by 2100)

Nuclear winter models:

100 nuclear weapons → nuclear winter (global cooling, crop failures, famine)
US-Russia exchange (thousands of weapons) → potential extinction-level event

Ecosystem collapse simulations:

Biodiversity loss, ocean acidification, soil degradation
Cascading failures in food systems

Economic system fragility:

Financial contagion, supply chain disruptions
Not existential alone, but amplifies other risks

Signal: Systems models show HIGH FRAGILITY (tipping points, cascades) or RESILIENCE (stable, self-correcting)

System 5: Institutional Preparedness

Pandemic response:

COVID-19 revealed gaps (slow response, lack of coordination)
Improvements: mRNA vaccines, better surveillance
But: Engineered pandemics could be worse

Nuclear arms control:

Treaties: NPT, START, INF (now expired)
Erosion of arms control (US-Russia tensions, China buildup)

AI governance:

Minimal governance currently (voluntary commitments, no binding treaties)
Proposals: International AI Safety Organization, compute governance

Biosecurity protocols:

Dual-use research oversight, gain-of-function research restrictions
But: Enforcement weak, garage biotech unregulated

Signal: Institutions are PREPARED (strong governance, coordination) or UNPREPARED (weak governance, gaps)

System 6: Public Awareness & Political Priority

Risk perception surveys:

Climate change: High awareness (70%+ concerned), but polarized
AI risk: Low awareness (most people unaware of existential risk)
Pandemics: High awareness post-COVID, but fading

Media coverage:

Climate: Extensive coverage
AI risk: Growing coverage (ChatGPT raised awareness)
Biosecurity: Minimal coverage (until pandemic)

Political priority:

Climate: High priority (Paris Agreement, net-zero commitments)
AI: Growing priority (EU AI Act, US executive orders)
Biosecurity: Low priority (underfunded)

Funding allocation:

Climate: Billions (but still insufficient)
AI safety: Millions (growing, but tiny compared to AI development)
Biosecurity: Underfunded relative to risk

Signal: Public/political awareness is HIGH (priority, funding) or LOW (ignored, underfunded)

System 7: Mitigation Efforts & Progress

Climate mitigation:

Renewable energy growth (solar, wind cost down 90%)
EV adoption accelerating
But: Emissions still rising, not on track for 1.5°C

AI safety research:

Growing field (alignment research, interpretability, robustness)
But: Safety research << AI capabilities research (imbalance)

Biosecurity:

Improved surveillance (genomic sequencing)
mRNA vaccine platforms (rapid response)
But: Dual-use research continues, garage biotech unregulated

Nuclear risk reduction:

Fewer warheads than Cold War peak (70K → 13K)
But: Modernization, new delivery systems, arms race resuming

Signal: Mitigation shows PROGRESS (risk decreasing) or INSUFFICIENT (risk stable or increasing)

System 8: Philosophical Frameworks

Longtermism:

Future generations matter morally (billions of potential future humans)
Existential risk reduction is top priority (preserves all future value)

Effective Altruism:

Focus on highest-impact interventions
Existential risk often neglected, high-leverage

Precautionary Principle:

When facing catastrophic risk with uncertainty, err on side of caution
Example: AI development—slow down if uncertain about safety

Existential risk taxonomy (Bostrom, Ord):

Extinction (humanity ends)
Unrecoverable collapse (civilization destroyed, can't rebuild)
Unrecoverable dystopia (locked into bad state)

Signal: Philosophical frameworks SUPPORT prioritization (longtermism, EA) or NEUTRAL (no strong ethical imperative)

Convergence-Based Risk Assessment

Case Study 1: Climate Change

System	Assessment	Signal	Confidence
Scientific Consensus	97% agreement, IPCC high confidence, warming >3°C catastrophic	HIGH RISK	0.90
Tech Trajectories	Emissions rising, tipping points approaching (2°C threshold)	ACCELERATING	0.80
Historical	Past climate shifts caused extinctions, civilizations collapsed	RISK REAL	0.70
Systems Modeling	IPCC models show tipping points, cascades (ice sheets, AMOC)	HIGH FRAGILITY	0.85
Institutional	Paris Agreement, but insufficient action, governance weak	UNPREPARED	0.60
Public Awareness	High awareness, political priority growing, billions in funding	HIGH	0.75
Mitigation	Renewables growing, but emissions still rising, not on track	INSUFFICIENT	0.65
Philosophical	Longtermism, precautionary principle support action	SUPPORT	0.80

Convergence Index: (0.90+0.80+0.70+0.85+0.60+0.75+0.65+0.80)/8 = 0.76

Interpretation: HIGH CONVERGENCE—climate change is severe threat (not extinction-level, but civilization threat), urgent action needed

Risk level: Catastrophic (not existential), high confidence

Priority: Top-tier (already high priority, but need more action)

Case Study 2: AI Existential Risk

System	Assessment	Signal	Confidence
Scientific Consensus	Divided—some experts 50% risk, others 5%, median ~10%	UNCERTAIN	0.55
Tech Trajectories	Rapid AI progress (GPT-4, AlphaFold), AGI timeline 2040-2060	ACCELERATING	0.75
Historical	No precedent for superintelligent AI (unprecedented risk)	UNKNOWN	0.50
Systems Modeling	Alignment problem unsolved, recursive self-improvement risks	HIGH FRAGILITY	0.70
Institutional	Minimal governance, voluntary commitments, no binding treaties	UNPREPARED	0.45
Public Awareness	Growing awareness (ChatGPT), but still low, underfunded	LOW	0.50
Mitigation	AI safety research growing, but << capabilities research	INSUFFICIENT	0.55
Philosophical	Longtermism, EA strongly support AI safety prioritization	SUPPORT	0.85

Convergence Index: (0.55+0.75+0.50+0.70+0.45+0.50+0.55+0.85)/8 = 0.61

Interpretation: MODERATE CONVERGENCE—AI risk is significant but uncertain, more research and governance needed

Risk level: Potentially existential, moderate-high uncertainty

Priority: High (underfunded relative to risk, need more investment)

Case Study 3: Nuclear War

System	Assessment	Signal	Confidence
Scientific Consensus	Nuclear winter models, 100+ weapons catastrophic, consensus strong	HIGH RISK	0.85
Tech Trajectories	Modernization, hypersonics, but fewer warheads than Cold War	STABLE	0.60
Historical	Near-misses (Cuban Missile Crisis, 1983), shows risk is real	RISK REAL	0.80
Systems Modeling	Nuclear winter models show civilization collapse, potential extinction	HIGH FRAGILITY	0.85
Institutional	Arms control eroding (INF expired), but some treaties remain (NPT)	WEAKENING	0.55
Public Awareness	Low awareness (post-Cold War complacency), underfunded	LOW	0.45
Mitigation	Fewer warheads, but modernization, arms race resuming	MIXED	0.60
Philosophical	Longtermism supports nuclear risk reduction	SUPPORT	0.75

Convergence Index: (0.85+0.60+0.80+0.85+0.55+0.45+0.60+0.75)/8 = 0.68

Interpretation: MODERATE-HIGH CONVERGENCE—nuclear war remains serious existential risk, complacency dangerous

Risk level: Existential (civilization collapse or extinction), moderate confidence

Priority: High (neglected post-Cold War, need renewed focus)

Existential Risk Hierarchy

Severe & High Confidence (CI > 0.70)

Climate Change (CI = 0.76): Catastrophic (not extinction), high confidence, urgent action
Nuclear War (CI = 0.68): Existential, moderate-high confidence, neglected

Action: Top priority, massive investment, international cooperation

Significant & Uncertain (CI 0.55-0.70)

AI Risk (CI = 0.61): Potentially existential, high uncertainty, underfunded
Engineered Pandemics (CI = 0.60): Emerging threat, growing risk, need governance

Action: High priority, invest in research, build governance, reduce uncertainty

Lower Priority or Overstated (CI < 0.55)

Asteroid Impact (CI = 0.45): Low probability, but high impact, some monitoring
Supervolcano (CI = 0.40): Very low probability, little we can do

Action: Monitor, but don't prioritize over higher-CI risks

Practical Application

For Researchers

High CI risks: Focus on mitigation (climate solutions, nuclear arms control)

Moderate CI risks: Focus on reducing uncertainty (AI safety research, biosecurity)

For Philanthropists

Funding allocation by CI:

50% to high-CI risks (climate, nuclear)
40% to moderate-CI risks (AI safety, biosecurity)
10% to low-CI or unknown risks (asteroids, unknown unknowns)

For Policymakers

High CI: Binding international agreements (climate, nuclear)

Moderate CI: Build governance frameworks (AI, biosecurity)

Conclusion: Convergence-Based Existential Risk Assessment

Convergence-based existential risk evaluation offers systematic framework for prioritizing humanity's long-term survival:

Multi-system integration: 8 independent risk assessment systems (scientific consensus, technological trajectories, historical precedents, systems modeling, institutional preparedness, public awareness, mitigation efforts, philosophical frameworks)
Risk CI: Quantifies threat severity and confidence
Risk hierarchy: Severe CI>0.70 (climate 0.76, nuclear 0.68), Significant CI 0.55-0.70 (AI 0.61, pandemics 0.60), Lower priority CI<0.55 (asteroids 0.45)
Case studies: Climate (CI=0.76 catastrophic high confidence), AI (CI=0.61 uncertain underfunded), Nuclear (CI=0.68 neglected)

The framework: